in-service performance in mobile media gateway · n ericsson offers in-service performance (isp)...

In-ServicePerformance In MobileMedia Gateway

Kalervo Kujala 30.09.2003S-38.310 Tietoverkkotekniikandiplomityöseminaari

Basic Information

n Thesis was written for Oy LM Ericsson ABn Professor Jorma Jormakkan Supervisor Juha Eloranta

Contents

n Problem descriptionn UMTS networkn Media Gatewayn In-Service Performancen ISP Statistics Tooln ISP Informationn Requirements / Conclusions

Problem Description

n Ericsson offers In-Service Performance(ISP) statistics from MSC (GSM-network)

n Same needed from MGW (UMTS-network)n ISP definition not formally defined

n How to get ISP statistics from the MGWnode and present it in an informative way?

UMTS network

n MGW - 3GPP R99 (+R4 functionality)n Both Release 99 and Release 4 presented

Release 99

Release 4

Mobile Media Gateway

n is in the border between different networksn transports circuit- and packet-switched

trafficn routing, switching and signalingn echo canceling and speech transcoding

In-Service Performance (ISP)

9s Availability Downtime/Year Examples1 90 % 36d 12h Personal Clients2 99 % 87h 36min Entry-level Businesses3 99.9 % 8h 46min Internet Service Providers, Mainstream Businesses4 99.99 % 52min 33s Data Centres5 99.999 % 5min 15s Carrier-grade Telcos, Medical, Banking6 99.9999 % 31.5s Military Defence System, Carrier-grade Goal

n Normally presented as ’9s’n Includes both negative and positive side of

performance

Literature

n ISO 9000, 9001 and 9004n ITU-T E.800, E.860 and E880n 3GPP Performance Management

TS 32.104 V3.6.0n QuEST TL 9000

ISO

n Customer Satisfactionn Performance measurements and

monitoring of the productn Conformance to determined requirementsn Gathered statistics stored to databasen Valid data analysis methods and

appropriate statistic techniques

ITU-T

n E.800: Terms and definitions for the Qos,network performance and dependability

n E.880: Service Level Agreement andmeasurement equations

n E.880: Guidelines to field data collection(both failures and successes) and storingto database

E.860 measurement equations

n Service Availability

n Service Unavailability

n Weighted Service Unavailability

%%100% UASA −=

%100_

_% ∗∑=

TimeActive

IntervalOutageUA

%100_

)_(% ∗∗∑=

TimeActive

SDFIntervalOutageUA 1<0 ≤SDF

3GPP Performance Management

n Beginning and end times of the serviceunavailability

n Transfer of the data to toher systems for laterprocessing (FTP)

n Results comparable to other vendors’ resultsn Measurement file format either ASN.1 with

binary encoding (BER) or XML (ExtensibleMarkup Language)

QuEST TL 9000

n Quality Excellence for Suppliers ofTelecommunications

n TL 9000 documents contain a consistentset of requirements and measurementsthat aim for faster, better and more cost-effective telecommunications services.

Measurement Rules

n Over 30 seconds’ duration of partial or total lossof primary functionality is counted to the SystemOutage.

n Over 15 seconds’ duration for a scheduled eventis counted.

n Outages due to natural disasters are notcounted.

n System Down Time (SDT) == System Outage

Measurement Equations

n Annualised Down Time

n Annualised Outage Frequency

N

PDT

m

ii∑

=∗= 112

N

mOF ∗= 12

Pi describes the duration of ith outageM is the total amount of outagesN describes the number of systems

in the service at the end of the monthTwelve is an annualisation factor.

ISP Statistics Tool

n Ericsson’s proprietary and internal tool forgenerating ISP statistics reports

n produces graphical reports for varioussystems developed in Ericsson

n For example, in the GSM system the MSCis such an element for which reports aregenerated

n intended for Ericsson’s design offices, customersupport, management and comparison toService Level Agreements

n valuable information from the operating nodesregarding fault tolerance and maturity of thesystem

n enables proactive actions in preventingdisturbances with similar fault situationsoccurring in different locations

n management can set ISP goals and follow themup

n comparison of the actual statistics to the SLAmade with the customer

Example graphs

Large Restarts / Exchange

0,000,050,100,150,200,250,300,350,400,45

2002

/06

2002

/07

2002

/08

2002

/09

2002

/10

2002

/11

2002

/12

2003

/01

2003

/02

2003

/03

2003

/04

2003

/05

Ave

rag

e n

o.

of

rest

arts

UPM

PLM

AUT

Downtime / Exchange

00,050,1

0,150,2

0,250,3

0,350,4

0,450,5

2002

/06

2002

/07

2002

/08

2002

/09

2002

/10

2002

/11

2002

/12

2003

/01

2003

/02

2003

/03

2003

/04

2003

/05

Do

wn

tim

e in

min

ute

s

CEF

PLM

AUT

UPM

6MRA

It should be noted that the figures in the graphs are fictional.

Equations used in ISP tool

n ISP tool’s Unavailability

n Six months’ rolling average

NTimeActive

SDFSDTUA

∗∗

= ∑_

)(% 0 >N 1<0 ≤SDF

6

6

1∑

−

==

i

i i

i

N

SDT

RA 0 >iN

SDT labelsn System Down Time (SDT) is presented in the ISP

statistics tool with four different labels: UnplannedManual (UPM), Automatic (AUT), Planned Manual(PLM) and Complete Exchange Failure (CEF).

n UPM means a manually initiated disturbance thatresults in SDT.

n AUT means automatically initiated disturbance.n PLM means manually initiated planned disturbance,

i.e. software and hardware upgrade.n CEF means an automatic or manual disturbance

event that results in prolonged SDT, which cannotbe recovered without a manual intervention

Conclusions about ISP tool

n No commercial softwares, which wouldfulfill the same functionality

n Expensive to build a new tool from scratchn Adabtable for generating reports for MGWn Need for modification, though

ISP information

n Two different logs in the MGW node:board log and availability log

n Which one is better and adequately fulfillsthe ISP definitions mentioned earlier

Board Log

n 9 entry places which are overwritten in around-robin manner

n Expresses only down eventsn Hard to modifyn Need to collect logs periodically from each

board

Availability Logn Format: XML version 1.0 (Extensible

Markup Language)n contains 1455 record places and the size

is approximately 1 MBn Works in a round-robin manner

<?xml version="1.0"?> <!DOCTYPEgreeting SYSTEM "hello.dtd"><greeting>Hello, world!</greeting>

n The log clients can run on either MPs orBPs.

n The clients can store node availability, PIU(Plug-In Unit) and service availabilityevents through Availability Log Interface(AVLI).

n The node and PIU events are Hardware(HW) events and the service events areSoftware (SW) events

Conclusions about logs

n Board Log is not adequaten Availability Log has many good qualities

n Dedicated ISP log too expensive

+ Up and Down events (for node not traffic)+ Enough space+ in accordance with 3GPP- need to estimate when traffic is up- SDT labels can only be divided to AUT, MAN and CEF- log gathering manual process- Starting unnecessary-/+ need to secure the transfer of the logs to ericsson’s facilities

Requirements / Conclusions

n ISP tool should be usedn Availability log is good enough for ISPn Rolling average need to be changed

n New graphs proposed

0≠n∑

∑

=

==n

ii

n

ii

N

SDTRA

1

1

New proposed graphsMP Restarts / node

0

0,5

1

1,5

2

2,5

3

2002

/06

2002

/07

2002

/08

2002

/09

2002

/10

2002

/11

2002

/12

2003

/01

2003

/02

2003

/03

2003

/04

2003

/05

Ave

rag

e n

o.

of

rest

arts

MAN

AUT

Node Restarts / node

0

0,2

0,4

0,6

0,8

1

1,2

1,4

1,6

1,8

2

2002

/06

2002

/07

2002

/08

2002

/09

2002

/10

2002

/11

2002

/12

2003

/01

2003

/02

2003

/03

2003

/04

2003

/05

Ave

rag

e n

o.

of

rest

arts

MAN

AUT

VMGW MP Restarts / node

0

0,2

0,4

0,6

0,8

1

1,2

1,4

2002

/06

2002

/07

2002

/08

2002

/09

2002

/10

2002

/11

2002

/12

2003

/01

2003

/02

2003

/03

2003

/04

2003

/05

Ave

rag

e n

o. o

f re

star

ts

MAN

AUT

System Down Time / node

0

0,5

1

1,5

2

2,5

3

3,5

2002

/06

2002

/07

2002

/08

2002

/09

2002

/10

2002

/11

2002

/12

2003

/01

2003

/02

2003

/03

2003

/04

2003

/05

Ave

rag

e S

DT

[m

in]

CEF

MAN

AUT

6MRA

Future Research

n In this thesis a framework for a solution was given.However, it was left to the future research a specificimplementation of the solution. A following list isprovided of the issues that can be more closelyresearched:

n The division of the restart label MAN to UPM and PLMby an analysis of the platform availability log

n The division of downtime into supplier-attributable andMNO-attributable downtime

n Formal definition of the concept In-Service Performancefor telecommunications systems

n Comparison of the format of ISP statistics betweendifferent equipment vendors

Questions?

Thank you for listening!